Information on Atmospheric Aerosol in OMI Measurements

1
Information on Atmospheric Aerosol in OMI
Measurements
Ben Veihelmann, Pieternel Levelt, Piet Stammes
and Pepijn Veefkind Royal Netherlands
Meteorological Institute (KNMI)
2
Overview

• Multi-wavelength aerosol algorithm
• Principal Component Analysis (PCA)
• Degrees of Freedom of Signal (DFS)
• Distinguish Aerosol Types
• Separate Aerosol Parameters
• Surface albedo
• Clouds

3
Multi-Wavelength Approach

• ? Best fitting aerosol model
• ? Information on Type, AOD, SSA, Size, Height?
• ? Surface reflectivity? Clouds?

Reflectance
Wavelength (nm)
4
Height Information from 477 nm ?
Effective cross section
O2-O2 Vertical
DistributionO2-O2
pressure2
Height
10-46 cm5/molec2
Aerosol layer
O2-O2 Density
Wavelength (nm)
O2-O2 Density
5
Principal Component Analysis

• Rlm Reflectance( l, Measurement )
• Covariance matrix RTR PTD P
• Principal components Pkl pk(l)
• Decompose R
• K Number of components necessary to
  reconstruct
• R with an error e lt enoise
• Number of Degrees of Freedom of Signal
• Set of K Weights W for a given measurement

K k1
Rlm S Wkm Pkl e
6
Synthetic Data for Reflectance R (l,model)

• Aerosol models (250)
• Biomass burning, Desert dust, Weakly
  absorbing
• - Aerosol Optical Depth (AOD) ... 0 5
• - Refractive Index various m n
  ik(l)
• gSingle Scattering Albedo (SSA) 0.8
  1
• - Size Distribution various
  bimodal
• - Height of layer 1-5 km
• Cloud models
• Geometries .. 8 m, 8 m0, 11 Df, ? 700
• Surface albedo (l) ... ocean, soil,
  vegetation

7
Principal Components
K k1
Rlm S Wkm Pkl e

• m0 0.6
• 0.9
• Df 20?

------- Principal component 1 ------- Principal
component 2 ------- Principal component 3 -------
Principal component 4
8
Degrees of Freedom of SignalSurface albedo
K k1
Rlm S Wkm Pkl e

• Soil/Veget.
• m0 0.6
• 0.9
• Df 20?

------- Soil ------- Vegetation ------- Ocean
9
Degrees of Freedom od SignalWavelength Band
Selection
K k1
Rlm S Wkm Pkl e

• Soil/Veget.
• m0 0.6
• 0.9
• Df 20?
• SNR1000

10

• Soil/Veget.
• m0 0.6
• 0.9
• Df 20?

- Biomass Burning x Desert Dust Weakly
Absorbing o Water Cloud D Ice Cloud
Lines connect points with const. ref. index,
height, size AOD 0, 0.1, 0.5 1.0, 2.5, 5.0
Weight 3
Weight 2
Weight 1
11
Distinguish Aerosol Types
- Biomass Burning x Desert Dust Weakly
Absorbing
12
Distinguish Aerosol Types AOD 0.5
- Biomass Burning x Desert Dust Weakly
Absorbing
AOD 5.0
2.5
1.0
0.5
0.1
0.1
0.5
1.0
5.0
2.5
13
Separate AOD and SSA
- Biomass Burning x Desert Dust Weakly
Absorbing
AOD
SSA
14
Surface Albedo Error /- 0.01
AOD lt 0.5
15
Surface Albedo Error /- 0.01
AOD lt 0.5
16
Distinguish Clouds 3 DFS
- Biomass Burning x Desert Dust Weakly
Absorbing o Water Cloud D Ice Cloud
17
Conclusions

• Aerosol multi-wavelength algorithm 20 bands 331 -
  500 nm
• Signal has 2 to 4 degrees of freedom
• number insensitive to surface (ocean, soil,
  vegetation)
• 477 nm band adds information
• O2-O2 absorption appears in 3rd PC and higher
• Distinguish Aerosol Types
• desert dust / weakly absorbing
• some ambiguity biomass burning
• Separate AOD and SSA for absorbing aerosol
• Surface albedo error minor impact for ADO 0.5
• Clouds can be distingiushed if number of DFS 3

18
Backup Material
19
Surface Albedo Error /- 0.01
impact minor for AOD 0.5
0.5
0.5
20
(No Transcript)
21
Degrees of Freedom of Signal Geometry
K k1
Rlm S Wkm Pkl e
Soil/Veget. m0 0.6 SNR1000
22
Outlook

• Non-spherical desert dust aerosol model
• Spheroidal shape approximation
• Validation
• AERONET ground based sunphotometer measurements
• Other satellite instruments PARASOL

23
Principal Component Analysis

• P transforms R to coordinate system with
  principal axes
• Number of dimensions can be reduced to K

Wkm Sl Pkl Rlm
Rlm Reflect.(l,model)
Wkm Weight (k,model)
K k1
Rlm S Wkm Pkl e
24
Nominal Retrieval PCA-Retrieval
LUT
decompose in PC
Rmodel(l) 250 models
Wmodel(k) 250 models
decompose in PC
AOT-interpol
AOT-interpol
Rmeas(l)
Wmeas(k)

• minimize
• Sl (Rmeas-Rmodel)2

• minimize
• Sk (Wmeas- Wmodel)2